Skip to content

Benchmark

ArmBench-LLM Armenian conversation (bleu, 0-shot)

Higher scores are better.

Models measured
27
Scores recorded
28
Sources
1
IQ weight
0.97

Score distribution

4.4
Mean
4.9
Median
1.9
Std dev
0.564
10th percentile
6.1
90th percentile
0.00169
Min
0.001697.4
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.97 for chat models, measured over 27 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude Sonnet 4· ArmBench-LLM Armenian data docs7.4
2Gemini 2.5 Flash· ArmBench-LLM Armenian data docs6.9
3Claude Sonnet 3.7· ArmBench-LLM Armenian data docs6.4
4GLM 4.7 Flash· ArmBench-LLM Armenian data docs6.1
5Gemma 3 27B IT· ArmBench-LLM Armenian data docs6
6GPT-5.4 mini· ArmBench-LLM Armenian data docs6
7Gemma 4 31B IT· ArmBench-LLM Armenian data docs5.7
8Qwen3.5 27B· ArmBench-LLM Armenian data docs5.4
9GPT-5.2 Prosetting not named by the source· ArmBench-LLM Armenian data docs5.3
10GPT-5.2 Prosetting not named by the source· ArmBench-LLM Armenian data docs5.3
11Grok 4.1 Fast· ArmBench-LLM Armenian data docs5
12GLM 5· ArmBench-LLM Armenian data docs4.9
13Gemini 3.1 Pro Preview· ArmBench-LLM Armenian data docs4.9
14Gemini 2.5 Pro· ArmBench-LLM Armenian data docs4.9
15Gemini 3 Flash· ArmBench-LLM Armenian data docs4.7
16Gemini 3 Pro Preview· ArmBench-LLM Armenian data docs4.7
17Gemma 3 12B IT· ArmBench-LLM Armenian data docs4.6
18GPT 5.4 Nano· ArmBench-LLM Armenian data docs4.5
19Grok 4 Fast· ArmBench-LLM Armenian data docs4.5
20Mistral Large 3· ArmBench-LLM Armenian data docs4.4
21Mistral Small 3.2 24B Instruct· ArmBench-LLM Armenian data docs4.4
22Gemma 3 4B IT· ArmBench-LLM Armenian data docs4
23Ministral 14B· ArmBench-LLM Armenian data docs3.5
24Qwen3 4B Instruct 2507· ArmBench-LLM Armenian data docs3.1
25Qwen3.5 35B A3B· ArmBench-LLM Armenian data docs0.564
26Qwen3 4B Base· ArmBench-LLM Armenian data docs0.542
27Grok 4.20· ArmBench-LLM Armenian data docs0.039
28Qwen3.5 9B· ArmBench-LLM Armenian data docs0.00169

Data credits: ArmBench-LLM by Metric-AI. Public numeric primary task results; no separate results licence asserted.